基于相似地理加权回归模型的中国区域人口死亡率影响因素空间异质性分析

Spatial Heterogeneity Analysis of Influencing Factors for Regional Population Mortality Rates in China Based on the Similarity and Geographically Weighted Regression Models

摘要: 人口死亡率与其影响因素之间的空间异质性是人口学和管理学等多个领域关注的重点。地理加权回归模型经常被用来处理地理数据中的空间异质性问题，然而该方法作为一种局部建模方法，只考虑了观测个体之间空间距离上的远近关系，没有考虑观测个体之间属性相似性。为了解决这一问题，本文使用相似地理加权回归模型，从人口结构、社会经济、医疗条件三个方面，分析2020年中国区域人口死亡率的空间格局及其影响机制。该模型在构造回归系数的局部加权最小二乘估计时，权函数同时考虑了地理邻近性与属性相似性，拟合效果更好。结果表明，人口结构、社会经济、医疗条件的解释变量对人口死亡率的影响随空间位置变化，具有明显的空间异质性，且各因素对于各地级行政区的人口死亡率具有不同程度和方向的影响。针对空间异质性特征，探讨了模型结果的稳健性，并提出了差异化的区域政策建议。

Abstract: Spatial heterogeneity between population mortality rates and their influencing factors is a key focus in fields such as demography and management science. Geographically Weighted Regression (GWR) models are frequently used to address spatial heterogeneity. However, as a local modeling technique, GWR only considers the spatial proximity between observation points and ignores their attribute similarity. To address this limitation, this study employs a Similarity-Geographically Weighted Regression (SGWR) model. Using data from three dimensions—population structure, socio-economic factors, and healthcare conditions—it analyzes the spatial patterns and influencing mechanisms of regional population mortality rates in China in 2020. This model constructs locally weighted least squares estimates for regression coefficients, incorporating both geographical proximity and attribute similarity into the weighting function, resulting in improved model fit. The results demonstrate that the influence of explanatory variables from population structure, socio-economic factors, and healthcare conditions on population mortality varies significantly across spatial locations, exhibiting clear spatial heterogeneity. Furthermore, each factor exerts varying degrees and directions of influence on mortality rates across different prefecture-level administrative divisions. Considering these spatial heterogeneity characteristics, the study examines the robustness of the model results and proposes tailored regional policy recommendations.